NZ Privacy Act Checklist for AI Delivery Teams

27 February 2026 · DataNAI

For implementation teams, privacy compliance must be embedded in architecture and operating procedures, not added at release time.

Delivery checklist

Data collection and purpose

  • Document why each data field is required for the workflow.
  • Remove fields that do not support the stated purpose.
  • Separate operational identifiers from model features where possible.

Access and security controls

  • Enforce role-based access boundaries.
  • Apply encryption in transit and at rest.
  • Keep an auditable log of access and policy changes.

Data quality and correction

  • Define processes for correcting inaccurate records.
  • Track source-of-truth ownership per dataset.
  • Monitor freshness and completeness for high-impact workflows.

Transparency and accountability

  • Provide internal documentation on model use in decision paths.
  • Define human review points for higher-risk outputs.
  • Record escalation and incident response ownership.

Retention and disposal

  • Set retention windows by data type and risk tier.
  • Implement deletion workflows that include derived stores.
  • Verify disposal controls during periodic audits.

Practical implementation pattern

Treat privacy as a build-time and run-time concern:

  1. include controls in design reviews,
  2. enforce release gates in CI/CD,
  3. validate controls in operational reviews.

References

Related next step

Turn this insight into a delivery plan for your team.